Field of the invention
[0001] This invention relates to wireless communication using multi-transmit multi-receive
antenna arrays, that is to say where both the transmitting and the receiving station
comprise an array of antenna elements. In cases where the antenna elements at a given
station may be used both for transmission and for reception, references herein to
a 'transmitter', a 'transmit antenna' a 'receiver' or a 'receive antenna' are to be
construed as references to the function that they are exercising during that operation.
Background of the invention
[0002] Wireless communication systems are assuming ever-increasing importance for the transmission
of data, which is to be understood in its largest sense as covering speech or other
sounds and images, for example, as well as abstract digital signals.
[0003] Currently proposed standards for wireless communication systems include the 3GPP
(3
rd generation Partnership Project) and 3GPP2 standards, which use Code Division Multiple
Access ('CDMA') and Frequency Division Duplex ('FDD') or Time Division Duplex ('TDD'),
the HIPERLAN and HIPERLAN2 local area network standards of the European Telecommunications
Standards Institute ('ETSI'), which use Time Division Duplex ('TDD') and the International
Telecommunications Union ('ITU') IMT-2000 standards. The present invention is applicable
to systems of these kinds and other wireless communication systems.
[0004] In order to improve the communication capacity of the systems while reducing the
sensitivity of the systems to noise and interference and limiting the power of the
transmissions, various techniques are used separately or in combination, including
space-time diversity, where the same data is transmitted over different transmit and/or
receive antenna elements, and frequency spreading, such as Orthogonal Frequency Division
Multiplex ('OFDM') where the same data is spread over different channels distinguished
by their sub-carrier frequency.
[0005] At the receiver, the detection of the symbols is performed utilising knowledge of
the complex channel attenuation and phase shifts: the Channel State Information ('CSI').
The Channel State Information is obtained at the receiver by measuring the value of
pilot signals transmitted together with the data from the transmitter. The knowledge
of the channel enables the received signals to be processed jointly according to the
Maximum Ratio Combining technique, in which the received signal is multiplied by the
Hermitian transpose of the estimated channel transfer matrix.
[0006] Two broad ways of managing the transmit diversity have been categorised as 'closed
loop' and 'open loop'. In closed loop signal transmission, information concerning
the transmission channels is utilised at the transmitter to improve the communication.
For example, the document Tdoc SMG2 UMTS-L1 318/98 presented to the ETSI UMTS Physical
Layer Expert Group describes operation of a Transmit Adaptive Array (Tx AA) FDD scheme
in which the dedicated channels are transmitted coherently with the same data and
code at each transmit antenna, but with antenna-specific amplitude and phase weighting.
The receiver uses pilots transmitted on the Common Channels to estimate separately
the channels seen from each antenna. The receiver estimates the weights that should
be applied at the transmitter to maximise the power received at the receiver, quantises
the weights and feeds them back to the transmitter. The transmitter applies the respective
quantised weights to the amplitudes and phases of the signals transmitted from each
transmit antenna of the array. Alternatively, in TDD systems, the channel state information
for weighting the signals applied to the downlink transmit antennas may be derived
from the uplink signals, assuming that the channels are equivalent, without transmission
of any specific channel or weighting information from the receiver to the transmitter.
[0007] Multi-Transmit-Multi-Receive ('MTMR') diversity schemes, where essentially the same
signal is transmitted in space-time diversity over the different combinations of transmit
and receive antenna elements, can provide significant gains in Signal-to-Noise Ratios
('SNR') and thus operate at low SNRs, enabling an increase in spectral efficiency
via the use of high order modulations. Alternatively, in multi-stream wireless communication
schemes, different signals can be transmitted between the transmit and receive antenna
element arrays enabling high spectral efficiency. However, multi-stream schemes of
this kind that have been proposed are viable only at high SNRs and require complex
receivers (for a N-Transmit and M-Receive antenna configuration, M must be greater
than or equal to N) in order to be able to extract the different transmitted signals
at the receiver.
[0008] An example of an open-loop multi-stream single user scheme is the Bell Labs layered
space-time ('BLAST') scheme described in an article by G. J. Foschini entitled "Layered
Space-Time Architecture for Wireless Communication in a fading Environment When Using
Multiple Antennas," Bell Laboratories Technical Journal, Vol. 1, No. 2, Autumn, 1996,
pp. 41-59.
[0009] A closed-loop alternative to the above scheme in which channel knowledge is used
at the transmitter for multi-stream transmission is described in an article by Mansoor
Ahmed, Joseph Pautler and Kamyar Rohani entitled "CDMA Receiver Performance for Multiple-Input
Multiple-Output Antenna Systems," Vehicular Technology Conference, Fall, Atlanta City,
Oct 2001. A schematic diagram illustrating the principle of this communication system
is shown in the accompanying Figure 1.
[0010] Such schemes are limited by compromises between diversity gain and spectral efficiency
and accordingly the range of operational SNRs is limited unless complexity is increased
or high modulation constellations (for example greater than 64 QAM) are used. The
present invention offers a substantial improvement in the compromise between diversity
gain and spectral efficiency.
Summary of the invention
[0011] The present invention provides a method of, and apparatus for, wireless communication
using multi-transmit multi-receive antenna arrays as described in the accompanying
claims.
Brief description of the drawings
[0012]
Figure 1 is a schematic diagram of a known generic multi-stream single user communication
system,
Figure 2 is a schematic diagram of a multi-stream communication system in accordance
with one embodiment of the invention, given by way of example,
Figure 3 is a graph illustrating the performance of the system of Figure 2 for different
spectral efficiencies,
Figure 4 is a graph illustrating the performance of the system of Figure 2 compared
with an open loop system with the same number of transmit antenna elements but different
numbers pf receive antenna elements, and
Figure 5 is a graph illustrating the performance of the system of Figure 2 compared
with an open loop system with the same numbers of transmit and receive antenna elements.
Detailed description of the preferred embodiments
[0013] Figure 1 of the drawings shows a known multi-stream wireless communication system
comprising a transmitter station 1 comprising a transmit antenna array 2 of
N transmit antenna elements and a receiver station 3 comprising a receive antenna array
4 of
M receive antenna elements. In the example illustrated in Figure 1,
N=M=2. A plurality of distinct data streams
x1 to
xF (
F = two in the example of Figure 1) are transmitted from the transmit antenna array
2 to the receive antenna array 4 and the data streams are weighted by respective complex
weighting coefficients
νn,f where
n is the n
th transmit antenna element and
f is the f
th data stream before being applied to the transmit antenna array. The distinct data
streams are separated and estimated at the receiver station in a linear or non-linear
receiver 5, to produce detected signals
s1 and
s2.
[0014] In the case shown in Figure 1, with
N=M=F=2, the propagation channel can be represented by a matrix
In the closed-loop system developed by Motorola and described in the article referred
to above by Mansoor Ahmed, Joseph Pautler and Kamyar Rohani, channel knowledge is
used at the transmitter for the multi-stream transmission. This scheme requires the
knowledge of the weight matrix,
V=[V1 V2], applied at the transmit antennas where
V1=[ν1,1 ν2,1]T and
V1=[ν2,1 ν2,2]T are two eigen-vectors of
HHH, (T and
H stand for transpose and conjugate transpose respectively). The inputs
n1 and
n2 shown in Figure 1 represent noise added to the signal channels.The noise is assumed
in the analysis below to be independent, identically distributed ('i.i.d.') complex-valued
Gaussian random values with variance σ
2 (AWGN noise). Finally y
1 and y
2 represent the respective received signals on the two antennas of the receive antenna
array 2.
[0015] It will be appreciated that the BLAST technique described in the article referred
to above by G. J. Foschini is equivalent to setting
ν1,1=ν2,2=1 and
ν1,2=
ν2,1=0, that is to say that each data stream is transmitted only on a single respective
transmit antenna element and no channel knowledge is used at the transmitter (open
loop).
[0016] It will also be appreciated that, in a conventional TxAA closed loop transmit diversity
scheme, a single stream is transmitted according to the eigenvector corresponding
to the maximum eigenvalue of
HHH, so that
V1=[ν1,1 ν2,1]T and V2=0. This is a closed loop single stream single user scheme whereas in the dual-stream
TxAA shown in figure 1 both eigenvectors
V1 and
V2 are used.
[0017] Analysis, in the context of High Speed Downlink Shared Channel (HS-DSCH) communication,
has arrived at two main conclusions regarding the dual-Stream Tx AA. The first conclusion
is that at low SNR (about -5dB), with turbo codes, H-ARQ and water-filling the closed-loop
dual-stream scheme can provide up to 50% increase in average throughput (Bits/Chip
Interval) when compared to the open loop dual-stream scheme. The second conclusion
is that for the closed-loop dual-stream scheme the performance (average throughput)
of a non-linear receiver is nearly the same as that with a linear receiver, that is
to say that the use of channel knowledge at the transmitter eliminates the need for
non-linear processing.
[0018] However, it has been found that single stream closed loop transmit diversity (Tx
AA) provides the best performance at mid and low SNR (-5 to 10 dB) and average throughput
of 0.5 to 3 bits/chip-interval. This is very important, given that the high SNR scenario
(>10 dB) conditions occur with low probability in cellular systems (especially CDMA
systems, for example).
[0019] The embodiment of the present invention shown in Figure 2 takes advantage of closed
loop transmit diversity while increasing the data rate by using multi-streaming. Similar
elements in Figure 2 to those of Figure 1 have the same numbering.
[0020] This embodiment of the invention is applicable generally where
F data streams are transmitted from respective sub-groups of the transmit antenna elements
at least one of which comprises a plurality of the transmit antenna elements. In a
preferred embodiment of the invention, each of the sub-groups of transmit antenna
elements has the same number
Nd of transmit antenna elements. In another embodiment of the present invention, the
sub-groups of transmit antenna elements have different numbers of transmit antenna
elements, each of the sub-groups comprising at least
Nd transmit antenna elements. Preferably, as in this embodiment of the invention, the
minimum number
Nd of transmit antenna elements in any sub-group is at least two. The use of more than
one antenna element in a sub-group improves the diversity of the communication for
that data stream, while the use of more than one sub-group improves the spectral efficiency
by transmitting different signals via the sub-groups. The choice of the configuration,
including the number of transmit antenna elements in each sub-group, and hence of
N and
Nd is an optimisation problem which can be formulated in the context of a given application
as a function of channel conditions and target performance, for example.
[0021] Depending on the target performance and functioning SNR, one can choose
Nd, and the number of groups in order to provide the needed diversity and spectral efficiency.
Moreover, one can also choose to set
Nd and the number of groups such that not all
N antennas are used, economising on calculation complexity at the receiver. This configuration
can be used in the case of good channel quality, thus high SNR and low target performance.
In one embodiment of the present invention, the numbers of antennas used in total
and in each sub-group and the value of
Nd are modified during operation of the system to adapt the choices to the current channel
conditions and target performance.
[0022] On the receiver side, this embodiment of the invention is applicable generally to
a number
M of receive antennas, where
M is greater than or equal to (
N/
Nd).
[0023] For the sake of simplicity, the preferred embodiment of the invention is illustrated
in Figure 2 for the case of 2 data-streams, 4 transmit antennas and 2 receive antennas
(
N=4,
Nd =2,
M=2). The multi-stream wireless communication system shown in Figure 2 comprises a
transmitter station 1 comprising a transmit antenna array 2 of two transmit antenna
elements and a receiver station 3 comprising a receive antenna array 4 of two receive
antenna elements. A linear or non-linear receiver 5 separates, decodes and demodulates
the signals received at the receive antenna array 4.
[0024] The elements of the transmit antenna array 2 are connected in two sub-groups 6 and
7. Two distinct data streams
x1 and
x2 are transmitted respectively from the transmit antenna sub-group 6 and from the transmit
antenna sub-group 7 to the receive antenna array 4. The data stream
x1 is weighted by complex weighting coefficients
ν1 and
ν1 before being applied to the two antenna elements of the sub-group 6 respectively
and the data stream
x2 is weighted by complex weighting coefficients
ν3 and
ν4 before being applied to the two antenna elements of the sub-group 7 respectively.
The distinct data streams are separated and estimated at the receiver station in a
linear or non-linear receiver 5, to produce detected signals
s1 and
s2.
[0025] In the case shown in Figure 2, with
N=M=Nd=2, the propagation channel can be represented by two matrices
and
where
hij represents the channel from the
ith transmit antenna element to the
jth receive antenna element.
[0026] The received signal vector can then be represented as follows:
where
and where the data streams are weighted by respective complex weighting coefficients
νn,f, n being the
nth transmit antenna element and
f the
fth data stream,
y1 and
y2 represent the respective received signals on the two antennas of the receive antenna
array 2, and
n1 and
n2 represent noise added to the signal channels, again assumed to be independent, identically
distributed ('i.i.d.') complex-valued Gaussian random values with variance
σ2 (AWGN noise).
[0027] Re-writing
Equation(1) in a vector form, we obtain that:
where
and the dimension of
Hequ is 2x2.
[0028] The estimated symbols (streams) at the output of a linear minimum mean square error
(MMSE) receiver are given by:
where
G = (
HequHHequ + σ2 I)
-1 HequH is the transfer function of the MMSE receiver,
I is the identity matrix and the superscript
H stands for the operation transpose conjugate.
[0029] For each stream the coefficients
V1 =[
ν1 ν
2]
T and
V2 =[
ν3 ν4]
T are chosen in order to maximize the received power P under unit norm constraint so
that the total transmit power is also normalized. The analytic solutions for
V1 and
V2, also called the eigenfilter solution (see for example chapters 4.4 and 4.5 of the
book "Adaptive filter theory" by Simon Haykin, published by Prentice Hall) are the
eigenvectors corresponding to the largest eigenvalues of the matrices
H1HH1 and
H2HH2, where
[0030] Using Equation (6), the two streams can be separated and estimated at the output
of the receiver, thus an increase in spectral efficiency of order 2 is obtained. In
addition, for coherent combining and diversity gain, the antenna coefficients V
1 and V
2 are chosen to maximize the receiver output power for each stream.
[0031] The performance of this embodiment of the present invention, referred to as multi-stream
transmit adaptive antenna ('M-TxAA') is shown in Figures 3, 4 and 5 for different
values of
N,
Nd and
M and spectral efficiencies, for the case
N=4,
Nd=2 and
M=2. The performance is evaluated in terms of un-coded bit error rate ('BER') as a
function of the ratio of transmit energy per bit to noise ('Tx Eb/No').
[0032] The results obtained with this embodiment of the invention are shown in figure 3
for different spectral efficiencies, given by different coding schemes: binary phase
shift key ('BPSK'), quadrature phase shift key ('QPSK'), and quadrature amplitude
modulation with 16 and 64 symbols per constellation ('QAM-16' and 'QAM-64').
[0033] Figure 4 shows a comparison between the performances of this embodiment of the present
invention (M-TxAA) and an open loop system ('OL') with the same number of transmit
antenna elements (four) and four receive antenna elements instead of this embodiment
of the present invention's two receive antenna elements. It will be seen that for
the given range of Tx Eb/No [6-20 dB], the performance is significantly improved when
M-TxAA is used compared to the multi-stream open loop scheme (BLAST). Furthermore,
for a given SNR and un-coded BER, (say 3e
-2 and 20 dB) M-TxAA achieves a bit rate of 12 Bits/Symbol (R=2x6) which is 50% higher
than the open loop multi-stream scheme. On the other hand, for a fixed bit rate and
a given un-coded BER (e.g. 8 bits/symbol and 3e-2) M-TxAA can operate at a SNR of
16.5 dB which is 3.5 dB less than the open loop multi-stream scheme. Note that for
these figures 3 and 4, only M=2 antennas is used at the receiver for M-TxAA, thus
resulting in a reduced mobile complexity, whereas the open loop multi-stream needs
at least M=4 receive antennas.
[0034] Figure 5 shows a comparison between the performances of this embodiment of the present
invention (M-TxAA) and an open loop system ('OL') with the same number (four) of transmit
antenna elements and of receive antenna elements. It will be seen that, for a given
spectral efficiency, e.g. 8bits/symbol, and a given un-coded BER, e.g. 3e
-2, M-TxAA can operate at a SNR of 10.0 dB, which is 10 dB less than the open loop multi-stream
scheme. Moreover, at the same un-coded BER of 3e
-2, for a bit rate 50% higher than the open loop (12 bits/symbol rather than 8 bits/symbol),
M-TxAA still can operate at an SNR of 14 dB, i.e., 4 dB lower.
[0035] The quantisation of the weights
V1=[ν
1 ν
2]
T and
V2=[ν
3 ν
4]
T can be performed as specified in the current 3GPP Rel'99 Closed loop transmit diversity
scheme. The elements ν
1 and ν
3 can be fixed to a constant power, and v
2 and v
4 are set to relative amplitude and phase (to ν
1 and ν
3 respectively). Thus only the two coefficients ν
2 and ν
4 need to be fed back which represents negligible additional overhead.
[0036] In the embodiments of the invention described above, the transmit antenna pairs (6)
and (7) form part of a single transmitter, that is to say that they are in the same
cell/sector. However it is also possible for them to form parts of two different sectors/cells
with which the mobile is in simultaneous communication during soft-handover/softer-handover.
Thus the mobile would receive, two separate streams from two different cells/sector
base-station transmitters.
[0037] The embodiments of the invention have been described above with specific reference
to the example where there are two transmit antenna sub-groups with two antenna elements
in each sub-group and two receive antenna elements. The adaptation of the above equations
to the more general case of
G sub-groups of transmit antenna elements, the sub-group
Gi comprising
Ni transmit antenna elements where
Ni ≥
Nd, and
M receive antenna elements gives the following equations (indicated for the flat-fading
case, the extension to the more general multi-path case being obtained by putting
corresponding vectors for the terms of the matrices):
[0038] Equation (1) becomes:
[0039] The values of
ui become (Equation 2)
with
i =
1,...,
G, note that the sum is for the first index only, that is if we represent
hindex1,index2, then index 1 is a sum as expressed above.
[0040] Equation 4 becomes:
with
[0041] The eigenfilter solution for
Vi (c.f. Equation 6) is then the eigenvector corresponding to the largest eigenvalue
of the matrix
HHi where: